J. Radiat. Res.

, 49, 83–91 (2008) Lecture Note

Cancer Risks and Low-Level Radiation in U. S. Shipyard Workers

Genevieve M. MATANOSKI*, James A. TONASCIA, Adolfo CORREA-VILLASEÑOR,

Katherine C. YATES, Nancy FINK, Elizabeth ELLIOTT,
Bruce SANDERS and Deborah LANTRY

US nuclear workers/Cancer mortality/Low dose radiation/Epidemiology.

The risks for four cancers, leukemia, lymphopoietic cancers (LHC), lung cancer and mesothelioma,
were studied in workers from shipyards involved in nuclear powered ship overhauls. The population rep-
resented a sample of all workers based on radiation dose at study termination. The final sample included
28,000 workers with ≥ 5.0 mSv, 10,462 workers with < 5.0 mSv and 33,353 non-nuclear workers. Nuclear
workers had lower mortality rates for leukemia and LHC than US white males but higher rates of lung
cancer and a significant five-fold excess of mesothelioma. Dose-dependent analyses of risks in the high
exposure group indicated that for each cancer the risk increased at exposures above 10.0 mSv. An internal
comparison of workers with 50.0 mSv exposures to workers with exposures of 5.0–9.9 mSv indicated
relative risks for leukemia of 2.41 (95% CI: 0.5, 23.8), for LHC, 2.94 (95% CI: 1.0,12.0), for lung cancer,
1.26 (95% CI: 0.9, 1.9) and for mesothelioma, 1.61 (95% CI: 0.4, 9.7) for the higher exposure group.
Except for LHC, these risks are not significant. However, the increasing risk with increasing exposure for
these cancers, some of which are known to be related to radiation, suggests that low-level protracted expo-
sures to gamma rays may be associated with these cancers. Other agents such as asbestos, which are com-
mon to shipyard work, may play a role especially in the risk of mesothelioma. Future follow up of the
population would identify bounds on radiation risks for this population for comparison with similar risks
estimated from other populations.

from estimated exposures at a single point in time based on

INTRODUCTION location of individuals in relation to the radiation source. For
the uranium miners, risks are derived only from estimated
The Japanese Atomic Bomb survivors have provided most exposures based on time spent in the mines. Scientists have
of the current knowledge concerning the risks from exposure questioned whether cumulative doses of gamma radiation
to external radiation.1–3) Risks from fractionated long-term based on one or a few high exposures versus multiple low
radiation exposures have been studied in underground ura- exposures over extended time periods as occur in occupa-
nium miners, but these exposures are limited to alpha tions are equivalent in terms of cancer risks. Recent research
particle radiation with lung cancer as the primary health out- has suggested that perhaps low doses of radiation may act
come.4–7) For the Japanese population risks are determined on cells in a different way than high dose exposure.8–12) In
addition, if doses based on low versus high dose rate expo-
*Corresponding author: Phone: +1-410-223-1866, sure are not equivalent, then examining the risks associated
Fax: +1-410-223-1867, with different external radiation exposure scenarios in pop-
E-mail: [email protected] ulations other than the Japanese would be important.
Department of Epidemiology, Johns Hopkins Bloomberg School of Public Approximately 30 years ago, the Department of Energy
Health, 111 Market Place, Suite 850, Baltimore, MD 21202-6709, U.S.A..
with the cooperation of the Navy undertook a study of ship-
This paper was presented on October 15, 2003, at the International Sympo-
sium “Health Effects of Low Dose Radiation on Cancer Risk” organized by yard workers in eight Navy and private shipyards that were
Low Dose Radiation Research Center, Central Research Institute of Electric involved in overhaul of nuclear vessels. The National Insti-
Power Industry, Tokyo. tute of Occupational Safety and Health (NIOSH) also under-
Lecture notes were assembled by K. Ishida, K. Sakai, and H. Tanooka and took a study of one of these yards. The initial NIOSH report
the paper rewritten by the author.
showed no risk of leukemia associated with radiation
The study described in this document has been reviewed and approved by
the Committee on the Review of Use of Data from Human Subjects of the exposures13) but a longer follow up has indicated that radia-
Johns Hopkins School of Public Heath. tion exposures are associated with an increased risk of leu-
doi:10.1269/jrr.06082 kemia mortality.14)

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

84 G. M. Matanoski et al.

reported results completed in the late 1980s include data

BACKGROUND from only seven shipyards and for only four selected cancers
that were examined in detail. These cancers were selected
The nuclear workers in shipyards differ from many radi- because of their suspected association with radiation and
ation-exposed workers in that their jobs are not directly asbestos exposure. The results summarize the data reported
related to working with a radiation source as do radiological to the U.S. Department of Energy in the early 1990s.
technicians for example. They are exposed incidentally to
radioactive materials when carrying out tasks related to their METHODS AND POPULATION
trades because the materials they handle have become radio-
active. For example, plumbers working on pipes and other Study population
materials that receive coolant water from the reactor are The initial population represented all employees in eight
exposed. Other workers performing the same job in other parts shipyards and consisted of about 800,000 workers, 700,000
of the ship remote from the reactor have little or no exposure. of whom were non-nuclear workers, as seen in Table 1. The
The source of exposure for shipyard workers is from neu- roster of workers was established from all records of
tron-activated corrosion products in the coolant system. The employees in each shipyard. Most workers in the non-
major effective radiation exposure is to cobalt-60 (half-life nuclear group were excluded because they had completed
5.3 years), which decays with 1 low energy beta emission employment before the start of nuclear overhauls, which
and 2 high energy gamma rays. The dosimetry database of represented the beginning of potential radiation exposure for
the Navy also includes as part of the monitoring program, a most jobs. Radiation exposed workers were defined as all
small number of radiographers and medical personnel who workers certified to work in areas having potential exposure
may have had different radiation exposures. to radioactivity at some time from the start of the nuclear
A second important characteristic of the shipyard popula- ship overhauls and who had an employment record and a
tion is that individual exposures have been measured from dosimetry record on the radiation database. Since women
the first overhaul. Members of the radiation safety program and African-Americans represented a very small number of
continuously monitor workers for compliance with both the population of nuclear workers, their data were excluded
wearing and positioning of badges or dosimeters. Therefore, from analysis. Follow up of all workers in one shipyard was
this shipyard population has measured exposures for each not complete at the time of analysis and so all data for that
individual for the entire period of overhauls. yard were omitted from the preliminary analysis. Non-nuclear
In summary, radiation exposure among shipyard workers is workers considered comparable to the nuclear workers were
associated with specific trades and areas worked within the defined as all persons employed at any time during overhauls
ship during overhaul. This exposure occurs at varying low of nuclear vessels in each shipyard who were not considered
dose levels over time and a comprehensive monitoring pro- certified to work in areas with potential radiation. Since this
gram provides dosimetry data for all individuals. Thus, this population of 77,000 nuclear and 117,000 non-nuclear
population provides a unique opportunity to study low dose workers represented a large group, follow up was limited to
exposure to gamma radiation over a prolonged period of time. a stratified random sample of the group. The final sample
The current article includes the early analysis of the mor- included all nuclear workers with lifetime exposures in the
tality of workers through January 1, 1982. These preliminary shipyard of 5.0 mSv or more at the time of death or end of

Table 1. Population for Study

Nuclear Workers Non-nuclear
≥ 5.0 mSv < 5.0 mSv
Initial Population 35,079 71,772 692,612
Study Population 28,060 49,449 117,718
Stratified Random Sample 28,000 10,462 33,353

Person-years of follow-up* 356,091 139,746 425,070

Median age at entry to follow-up 34.3 33.6 33.4

Median year at entry to follow-up 1968.1 1967.0 1967.3
* Unconfirmed deaths omitted from analysis.
Includes data from seven shipyards only.

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

 Cancer Mortality in Shipyard Workers 85

follow up on January 1982 (28,000), about 20% of nuclear ification variable to control for general industrial exposures
workers with lifetime exposures below that dose (10,462), among the three exposure groups. The hazard index score
and about 30% of non-nuclear workers (33,353). Additional was developed for each job through a unique procedure that
exclusions included military personnel and workers with ranked all jobs based on the overall exposure to hazardous
employment for less than 1 year.13) substances for that job. A team of industrial hygienists with
experience in the shipyard environment were polled until
Exposure assessment they reached a consensus agreement on the hazard score for
Nuclear ship overhauls began at different times in each of each job. For example, the jobs that represented three or
the eight shipyards from 1957 through 1967 but similar radi- more percent of the workers and were included in the high
ation monitoring programs were started in each shipyard rank level were welder, painter and pipefitter. In this scheme
with initial overhaul. The monitoring techniques to deter- a job hazard index is not determined from a single agent but
mine exposures of individuals differed during the period of from many possible hazardous agents. This stratification
follow up. Exposures were measured initially using film variable was required for the entire population of nuclear
badges with open and closed windows to record both beta and non-nuclear workers for purposes of sampling. The only
and gamma radiation. However, the population rarely job records that could be abstracted efficiently in order to
received recordable beta radiation. The badges were read assign a hazard score from such a large number of records
monthly. The minimum detectable effective dose was esti- (about 200,000) was the last job held. Thus, each individual
mated to be 0.1 mSv. In 1973 to 1974, all shipyards convert- was ranked based on his last job. Since most shipyard jobs
ed to thermal luminescent dosimeters (TLDs) that were read represent skilled trades, this classification procedure was
daily. The new procedure led to a reduction in all annual deemed appropriate since most workers remain in one trade
doses due to the limitation of measuring low doses in the throughout their shipyard employment. Pre-nuclear work
minimum detection range when cumulated doses are frac- periods represented the time employed prior to the start of
tionated into daily rather than monthly readings. The actual dosimetry records for individual nuclear workers. In order to
recorded radiation exposure for each worker was used to balance the groups for the duration worked before starting
represent the effective dose of cobalt 60 without adjustment nuclear work, a pseudo pre-nuclear period was developed for
for differences in potential detection level with different each non-nuclear worker by creating a subgroup of these
devices, differing procedures over time or other potential workers that were similar to the ≥ 5.0 mSv subgroup for the
sources of exposure to radiation such as medical x-rays. The other four variables. Then the non-nuclear workers were
radiation records were obtained from a database available assigned a pseudo pre-nuclear work variable, which repre-
from the Navy. The information was validated on a random sented a duration of “pre-nuclear” work that was the same
sample of individual records using the workers medical or greater than the distribution of that variable in the com-
records where dose was recorded independently. The effec- parable nuclear reference group. Since 40 percent of nuclear
tive dose was used to define the samples and, thus, was used workers were recruited into the radiation program after five
in the analysis for consistency in reviewing all cancers. or more years of employment, inclusion of the “pseudo pre-
Organ doses will be considered for any future analysis. nuclear work” time variable assured that the non-nuclear
The yards require constant monitoring of each worker for worker group had the same long employment prior to radi-
compliance in wearing dosimeters. They also conduct con- ation start as did the nuclear workers. This avoided any
tinuous air monitoring programs to detect air releases of “survival bias” that might occur in nuclear workers related
radiation. Routine surveillance procedures include external to the fact that they were actively working at recruitment
and internal reviews to validate the accuracy of dosimeter into radiation work.
measurements using standard exposures.15) The annual and As seen in Table 2, the three exposure categories were not
cumulative measurements of radiation exposures from mon- similar based on the five strata characteristics before selec-
itoring were available for each nuclear worker from the start tion of the sample. Non-nuclear workers had increased pro-
of overhaul to end of follow up on January 1, 1982. portions in the older and younger age groups, more recent
years of hire, jobs with lower hazard indices and shorter
Study sample durations worked. After random selection of the samples
The study sample was selected to be similar in distribu- from the nuclear worker group at < 5.0 mSv and the non-
tion to the characteristics of the ≥ 5.0 mSv group in regard nuclear workers, all three groups were similar in all charac-
to birth year, hire year, total duration worked, duration of teristics (Table 3).
pre-nuclear work and job hazard index. Simultaneous adjust- Although jobs were not directly compared in the three
ment for these time factors during the sample selection groups, all nuclear compared to non-nuclear workers initial-
resulted in a carefully balanced population in regard to time ly did have some differences in the distribution of jobs based
of exposures. The job hazard index was not intended to con- on last job title. Jobs were divided into 49 occupational title
trol for specific shipyard exposures but was used as a strat- codes as well as 21 “prefix” codes, the latter representing a

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

86 G. M. Matanoski et al.

Table 2. Characteristics of Study Population by Sampling Table 3. Characteristics of Random Sample by Strata, Birth-
Strata, Birthyear, Year of Hire, Job Hazard Index, Duration of year, Year of Hire, Job Hazard Index, Duration of Pre-Nuclear
Pre-nuclear Work and Total Duration Worked* Work and Total Duration Worked
NW ≥ 5.0 NW < 5.0 NNW NW ≥ 5.0 NW < 5.0 NNW
(N = 28,542) (N = 49,635) (N = 119,179) (N = 28,542) (N = 10,462) (N = 33,353)
Variable No. % No. % No. % Variable No. % No. % No. %
Birthyear Birthyear
<1920 5240 18 11967 24 26984 23 <1920 5240 18% 1920 18% 5861 18%
1920–1929 6298 22 9702 20 16658 14 1920–1929 6298 22% 2307 22% 7195 22%
1930–1939 6486 23 9959 20 15979 13 1930–1939 6486 23% 2382 23% 7802 23%
1940–1949 8345 29 13994 28 30958 26 1940–1949 8345 29% 3059 29% 9932 30%
1950+ 2173 8 4013 8 28600 24 1950+ 2173 8% 794 8% 2563 8%

Year of Hire Year of Hire

<1950 5435 19 11185 23 21627 18 <1950 5435 19% 1976 19% 5860 18%
1950–1959 5977 21 9771 20 11294 10 1950–1959 5977 21% 2152 21% 6221 19%
1960–1969 12048 42 20078 41 36079 30 1960–1969 12048 42% 4462 43% 15035 45%
1970+ 5082 18 8601 17 50179 42 1970+ 5082 18% 1872 18% 6237 19%

Job Hazard Index Job Hazard Index

Missing 1440 5 2152 4 3940 3 Missing 1440 5% 547 5% 1727 5%
Low 5038 18 13601 27 26238 22 Low 5038 18% 1929 18% 7534 23%
Medium 13047 46 23585 48 59452 50 Medium 13047 46% 4783 46% 14918 45%
High 9017 32 10297 21 29549 25 High 9017 32% 3203 31% 9174 28%

Duration of Pre-Nuclear Work Duration of Pre-Nuclear Work

0- years 10685 37 16413 33 N/A N/A 0- years 10685 37% 3874 37% 12871 37%
2- years 6326 22 10447 21 N/A N/A 2- years 6326 22% 2308 22% 7313 22%
5- years 4348 15 7293 15 N/A N/A 5- years 4348 15% 1565 15% 5003 15%
10- years 4272 15 6763 14 N/A N/A 10- years 4272 15% 1238 12% 4633 14%
20+ years 2911 10 8719 18 N/A N/A 20+ years 2911 10% 1477 14% 3533 11%

Total Duration Worked Total Duration Worked

0- years 870 3 2709 5 13531 11 0- years 870 3% 537 4% 2145 6%
2- years 2822 10 8923 18 38350 32 2- years 2822 10% 1991 19% 7561 23%
5- years 5554 19 10318 21 26367 22 5- years 5554 19% 2089 20% 7702 23%
10- years 8716 31 11454 23 15572 13 10- years 8716 31% 2650 26% 7428 22%
20+ years 10580 37 16231 33 25359 21 20+ years 10580 37% 3195 31% 8517 26%
*Includes data from eight shipyards

skill or responsibility level (e.g. foreman). The differences in of “prefix” codes indicates that, in general, the nuclear work-
the distribution of jobs between nuclear and non-nuclear ers have a higher percentage of foremen, inspectors and
workers were small and would have been partially controlled planners among these titles compared to non-nuclear work-
by the stratification on the hazard index (Table 4). A review ers whereas the latter group have more “worker” and “help-

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

 Cancer Mortality in Shipyard Workers 87

Table 4. Distribution of Selected Last Occupational Title outcomes were determined from underlying causes of death
Codes* for Workers in Four Shipyards Based on Radiation Sta- listed on death certificates with confirmation of the diag-
tus noses of selected cancers through review of hospital records
Non-nuclear and cancer registry databases. The cancers in the current
Nuclear Workers Workers analysis do not reflect the results of these hospital records
(N = 24130) (N = 125925) Significant searches. Since the U.S. data used for comparison consisted
of death certificate information only, the same data had to be
Job Title* N % N % at p < = 0.05
used for causes of death for the shipyard workers in order to
5 Electrician 1942 8 9087 7 NS be comparable. Hospital records were used only to verify the
8 Engineer 2418 10 8075 6 S accuracy of the death certificates. For these cancers this ver-
ification process confirmed that virtually all death certificate
23 Laborer 200 1 6120 5 S
results were accurate. The study team obtained death certif-
25 Machinist 3090 13 16215 13 NS icates on over 99 percent of the suspected deaths from the
30 Nuclear Engineer 989 4 1815 1 S four specific cancers.
35 Pipefitter 2777 12 7795 6 S
37 Rigger 1103 5 4874 4 NS
Analysis
Two independent analyses were performed on the data.
42 Shipfitter 1426 6 10394 8 S First, a categorical analysis was performed using three life-
47 Welder 1348 6 11306 9 S time cumulative dose groups defined as cumulative dose at
88 Administrative 323 1 4894 4 S death or study termination corrected by age and calendar
time to US white male rates. A second dose sensitive anal-
*Specific job titles are included if they represented four percent or
more of jobs for either nuclear or non-nuclear worker. ysis using the Life Table method16) was performed on work-
ers with lifetime exposures of 5.0 mSv or more. For this
analysis, the risk of cancer deaths for each individual at each
er” classifications. (Data are not shown.) As shown in Table dose they experienced at a specified age and calendar time
4, among the specific job titles that represent at least four period was compared to the risk in the 5–9 mSv reference
percent or more of the distribution of job titles of either population using the same adjustment methods. This time
nuclear or non-nuclear workers, three were significantly dependent analysis results in all individuals with high cumu-
higher in nuclear and four in non-nuclear workers with the lative doses adding person-years to all lower dose groups
remaining three jobs equally distributed in the two groups. included in the time dependent analysis although they would
The most important differences may be the higher propor- not have an event at the lower doses. The group with < 5
tion of pipefitters among nuclear workers and the higher mSv could not be used as a reference because that group was
proportion of shipfitters and welders among non-nuclear excluded from the time dependent analysis. Since the cate-
workers. Thus, even after balancing the populations by job gory of < 5.0 mSv was a sample of all workers with a final
hazard index, the nuclear workers could differ from non- dose of that level and since the multistrata sampling was
nuclear workers in regard to exposure to other agents com- complicated making it impossible to reconstruct the original
mon in shipyards. population from the sample, this group could not be includ-
ed as part of the dose, age and calendar time dependent
Follow up of population analysis. Adding person-years from the 5 mSv or more
Multiple data sources were searched to determine the vital population to a sample of the < 5 mSv group would have
status of each worker including Social Security Administra- inflated the person-years in the < 5 mSv exposure group and
tion files, Civil Service Administration files, Health Care artificially lowered the risk in this group. The risks were ana-
Financing Administration (HCFA) files, Virginia mortality lyzed with lagging for different exposure periods. This
files and the National Death Index. All of these files could adjustment excludes any radiation exposure during a speci-
determine whether a worker was deceased. In addition, fied “lag period” from the worker’s exposure for the person
Social Security, HCFA, and Civil Service Administration years or events accrued in that period. Thus, the cumulative
had separate files that could determine whether the enrollee person-years and any events occurring at any point in the
was alive. All shipyard workers who were current employees follow up are attributed to the cumulative radiation dose to
at any time during the final year of the study were classified that time minus any dose accrued during the lag period. The
as alive if there was no confirmation of a death from a death lag periods included no lag and 2, 5, 10, and 15 years. The
certificate. Vital status, either living or deceased, was deter- short lag period represented the time when first cases of leu-
mined on all but four percent of the workers. Follow up was kemia reportedly have appeared after exposures to a known
not complete for one shipyard at study termination and all leukemogen. The longer lag periods were included to repre-
data on those workers are excluded from the analysis. The sent possible latency periods of solid tumors. However, the

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

88 G. M. Matanoski et al.

number of workers with 10 or more years of follow up was Table 6. Age-adjusted Leukemia Mortality by Radiation
limited since the median follow up period for all workers Group
was 13 years. Therefore, the analysis reported includes only External Comparison to US Male Population:
a lag of two years for leukemia and five years for all other
Non-nuclear
cancers as well as the analysis with no lag. Nuclear Workers Workers

RESULTS ≥ 5.0 mSv < 5.0 mSv

Deaths 21 4 29
Four cancer sites were selected for initial analyses SMR 0.91 0.42 0.97
because of their potential association with either radiation or
95% CI (0.56,1.39) (0.11,1.07) (0.65,1.39)
asbestos exposure: leukemia, lymphoma and hematopoetic
cancers, lung cancer and mesothelioma. For each cancer, the Internal Comparison:
upper portion of the Tables 5–9 represents the overall cate-
Workers No Lag Lag 2 years
gorical results by lifetime radiation exposure status and the
lower portion represents the calendar time, age and dose Relative Relative
dependent Life Table analysis of the complete 5.0 mSv or Nuclear ≥ 5.0 mSv Risk 95% CI Risk 95% CI
more population.
5.0-* 1.00 – 1.00 –
The all cause mortality ratios, as shown in Table 5, indi-
cate that the risk of death in shipyard workers is equal to or 10.0-* 2.63 (0.6,24.2) 3.05 (0.7,27.8)
lower than that of US white males. However, only in the two 50.0-* 2.41 (0.5,23.8) 2.29 (0.4,23.2)
nuclear categories are the risks significantly lower than the
Nuclear < 5.0 mSv 1.02 (0.2,11.3) 1.07 (0.2,11.9)
general population. The relative risks for all outcomes are
higher in both nuclear workers at < 5.0 mSv and non-nuclear Non-nuclear 2.37 (0.6,20.5) 2.54 (0.6,21.9)
workers than in the reference category of nuclear workers
*Dose dependent analysis (Reference 5.0–9.9 mSv)
with doses of 5.0 to 10.0 mSv . In the dose dependent anal- Adjusted for age and calendar time by indirect adjustment
ysis with a lag of five years, the risk increases with increas-
ing dose. This increase, even at the highest dose, is not

Table 5. Age-adjusted All Cause Mortality by Radiation Table 7. Age-adjusted Lymphatic and Hematopoietic Can-
Group cer Mortality by Radiation Group
External Comparison to US Male Population: External Comparison to US Male Population:

Non-nuclear Non-nuclear
Nuclear Workers Workers Nuclear Workers Workers

≥ 5.0 mSv < 5.0 mSv ≥ 5.0 mSv < 5.0 mSv
Deaths 2215 973 3745 Deaths 50 13 84
SMR 0.76 0.81 1.00 SMR 0.82 0.53 1.1
95% CI (0.73, 0.79) (0.76, 0.86) (0.97, 1.03) 95% CI (0.61, 1.08) (0.28, 0.91) (0.88, 1.37)

Internal Comparison: Internal Comparison:

Workers No Lag Lag 5 years Workers No Lag Lag 5 years
Relative Relative Relative Relative
Nuclear ≥ 5.0 mSv Risk 95% CI Risk 95% CI Nuclear ≥ 5.0 mSv Risk 95% CI Risk 95% CI
5.0-* 1.00 – 1.00 – 5.0-* 1.00 – 1.00 –
10.0-* 1.10 (0.89,1.23) 1.08 (0.96,1.22) 10.0-* 3.23 (1.1, 12.6) 5.84 (1.5, 50.8)
50.0-* 1.03 (0.91,1.16) 1.13 (0.99,1.30) 50.0-* 2.94 (1.0, 12.0) 5.37 (1.2, 48.7)

Nuclear < 5.0 mSv 1.13 (1.01,1.26) 1.15 (1.02,1.30) Nuclear < 5.0 mSv 1.71 (0.57, 7.2) 2.95 (0.6, 27.4)

Non-nuclear 1.39 (1.26,1.53) 1.39 (1.24,1.55) Non-nuclear 3.58 (1.3, 13.5) 5.63 (1.5, 47.5)
*Dose dependent analysis (Reference 5.0–9.9 mSv) *Dose dependent analysis (Reference 5.0–9.9 mSv)
Adjusted for age and calendar time by indirect adjustment Adjusted for age and calendar time by indirect adjustment

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

 Cancer Mortality in Shipyard Workers 89

Table 8. Age-adjusted Lung Cancer Mortality by Radiation significant.

Group The leukemia risk by lifelong exposure category indicates
External Comparison to US Male Population: that the highest mortality ratio compared to U.S. white males
occurs in non-nuclear workers as seen in Table 6. Although
Non-nuclear
Nuclear Workers Workers
all ratios indicate risks below that of the general population
of US white males, comparison of the low and high radiation
≥ 5.0 mSv < 5.0 mSv exposure in categorical groups indicates a leukemia mortal-
Deaths 237 98 306 ity that is two times higher in the group with highest lifetime
SMR 1.07 1.11 1.15 exposures compared to the low exposure group. In the dose
dependent analysis shown in the lower portion of the table,
95% CI (0.94, 1.21) (0.90, 1.35) (1.02, 1.29)
there is a suggestion of an increase in risk with dose but the
Internal Comparison:
relative risks in each dose group are non-significant due to
small numbers. The risk of leukemia for workers with doses
Workers No Lag Lag 5 years
of 50 mSv or more is 2.41 (95% CI: 0.5, 23.8) times higher
Relative Relative than the reference category of 5.0–9.9 mSv. The numbers by
Nuclear ≥ 5.0 mSv Risk 95% CI Risk 95% CI cell type are too small for analyses. Myeloid leukemias rep-
5.0-* 1.00 – 1.00 –
resent 60 percent of all cases with identified types.
The overall mortality from lymphatic and hematopoetic
10.0-* 1.08 (0.8, 1.6) 1.05 (0.7, 1.6)
cancers (LHC) including leukemia shown in Table 6 indi-
50.0-* 1.26 (0.9, 1.9) 1.36 (0.9, 2.1) cates a significantly low LHC death rate for those exposed
at < 5.0 mSv compared to US white males with the pattern
Nuclear < 5.0 mSv 1.16 (0.8, 1.7) 1.25 (0.9, 1.9)
of SMRs by category that is similar to that for leukemia. The
Non-nuclear 1.21 (0.9, 1.7) 1.22 (0.9, 1.7) ratio for those in the ≥ 5.0 mSv category is higher than the
lowest radiation exposure group. In the dose dependent anal-
*Dose dependent analysis (Reference 5.0–9.9 mSv)
Adjusted for age and calendar time by indirect adjustment ysis, the relative risk at 50 mSv or more is significantly high-
er (RR = 2.94; 95% CI: 1.0, 12.0) than risks at 5.0 to 10.0
mSv. That difference is even more apparent when the expo-
sures in the last 5 years of follow up before a specific
Table 9. Age-adjusted Mesothelioma Mortality by Radiation calendar time or event (lag 5 years) are omitted (RR = 5.37;
Group 95% CI: 1.2, 48.7).
The lung cancer standardized mortality ratios using the
External Comparison to US Male Population:
comparison US population for the life long exposure cate-
Nuclear Workers Non-nuclear gories indicate little difference in ratio between the three
Workers
groups as shown in Table 8. All ratios are above 1.00 but
≥ 5.0 mSv < 5.0 mSv only non-nuclear workers have a significantly higher mortal-
Deaths 18 8 10 ity ratio from lung cancer compared to US white males. In
the dose dependent analysis, the risk of lung cancer mortal-
SMR 5.11 5.75 2.41
ity is about 30 percent higher for those receiving 50 mSv or
95% CI (3.03, 8.08) (2.48, 11.33) (1.16, 4.43) more exposure compared to 5.0 to 10.0 mSv (RR = 1.26;
95% CI: 0.9,1.9) and there is a suggested dose response for
Internal Comparison:
both lag-periods. For the 277 lung cancers with hospital
Workers No Lag Lag 5 years record reviews, data on morphology indicated 36 percent
Relative Relative had squamous cell carcinoma, 20 percent had adenocarcino-
Nuclear ≥ 5.0 mSv Risk 95% CI Risk 95% CI
ma, 15 percent had small cell carcinoma and 18 percent had
an undifferentiated cell type with smaller percentages of oth-
5.0-* 1.00 – 1.00 –
er cell types.
10.0-* 1.21 (0.3, 9.1) 1.24 (0.3, 7.2) Mesothelioma was examined because radiation workers’
50.0-* 1.61 (0.4, 9.7) 1.50 (0.3, 9.2) jobs may involve exposures both to asbestos and radiation.
The mortality from mesothelioma based on life long catego-
Nuclear < 5.0 mSv 1.45 (0.4, 8.5) 1.31 (0.3, 7.8) ries of radiation exposure indicate that shipyard workers in
Non-nuclear 0.61 (0.2, 3.4) 0.57 (0.1, 3.3)
all groups have 2.4 to 5.8 times higher mortality compared
to US white males and that SMRs are 2.1 to 2.4 times higher
*Dose dependent analysis (Reference 5.0 – 9.9 mSv)
in the radiation exposed groups compared to the non-
Adjusted for age and calendar time by indirect adjustment
exposed (Table 9). The dose dependent comparison of risks

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

90 G. M. Matanoski et al.

suggests that the risk of mesothelioma increases with as well as a work exposure factor termed the hazard index.
increasing radiation in the ≥ 5.0 mSv group. The dose cate- However, that hazard factor was not meant to control for
gory of 50.0 mSv or more has a RR = 1.6 (95% CI: 0.4, 9.7) specific exposures. Therefore, the increased risk of mesothe-
compared to the workers with 5.0 to 10.0 mSv exposures. lioma and lung cancer could have been related to confound-
None of these observations are significantly different from ing exposures to asbestos and for lung cancer, differences in
1.0 (or no excess risk) due to small numbers in each catego- smoking by radiation exposure also could have played a role
ry. The increases could be related to differences in asbestos in the observed risks. Unfortunately, the smoking habits of
exposure that may be correlated with the radiation dose all individuals in the study were not available. However, a
although there is no apparent reason to suggest that this survey of one shipyard suggested no difference in the age-
should occur. adjusted rates of ever smokers in nuclear and non-nuclear
workers and a lower rate of current smokers in the nuclear
DISCUSSION groups. Thus, if all shipyards were similar to the one sur-
veyed, smoking could have explained the difference in the
Workers involved with overhaul of nuclear powered ships higher lung cancer risks in non-nuclear versus nuclear work-
represent an occupational group exposed intermittently to ers. However, smoking differences are unlikely explanations
low-level gamma-ray exposure from cobalt 60 while per- for the increased risks observed in the higher radiation
forming their usual jobs. The mortality analysis based on exposed group versus the lower dose reference groups.
lifetime radiation exposure shows that these nuclear workers The data on jobs of workers indicate a higher proportion
with cumulative exposures both below and above 5.0 mSv of pipefitter jobs in nuclear versus non-nuclear workers.
have lower risks of leukemia, lung cancer and LHC than Since asbestos is one hazardous agent associated with pipe-
non-nuclear workers. The radiation exposed workers are fitting work, one might expect higher rates of mesothelioma
selected into the program from the group of all shipyard in nuclear compared to non-nuclear workers and that is the
workers primarily on the basis of work performance since case (Table 9). However, although jobs may have differed
one of the shipyard’s aims is to complete an overhaul of the between nuclear and non-nuclear workers, it is less likely
reactor area in the shortest time period possible to reduce the that the jobs differed between workers at the various dose
total radiation exposure per overhaul. The workers also must levels. Therefore, the increases in risk by radiation dose,
complete a physical exam, radiation training and certifica- even though not significant, suggest that radiation may be
tion prior to being allowed to work in the potentially playing a role. Further investigation of the joint effects of
exposed areas. Thus, selection bias may explain the overall asbestos and radiation on the risk of mesothelioma has been
lower SMR of these cancers in nuclear versus non-nuclear completed on the total study population by one of the co-
workers since healthier and more efficient workers may be investigators.17) The results suggest that each agent may
selected for radiation training and certification. Mesothelio- independently increase the risk of this cancer. Further adjust-
ma risks that are known to be associated with asbestos ment for possible confounding occupational exposures
exposure actually show higher SMRs in nuclear versus non- should be included as part of future studies of this popula-
nuclear workers suggesting there may be differences in tion. Stratification on the general hazard index in the study
exposure to confounding agents, especially asbestos, in sample should not interfere with the ability to examine spe-
nuclear work. For example, Table 4 suggests a higher pro- cific agents of interest.
portion of pipefitters who have potential asbestos exposure The only results that conflict with this apparent dose
among nuclear compared to non-nuclear workers. response effect from radiation are that for many of these
The dose dependent data analyses suggest that doses of cancers, leukemia, lung cancer and LHC, the non-nuclear
radiation of 10.0 mSv or more are associated with an workers have as high a risk as the highest radiation dose. As
increased risk of leukemia, LHC, lung cancer and mesothe- suggested, a probable explanation for these findings is that
lioma compared to workers with doses of 5.0 to 10.0 mSv. selection of workers for the radiation program may intro-
For each cancer the risks appear to increase with increased duce a bias that is reflected in inherently low risks of cancer
dose but few values except for LHC show significant differ- in nuclear versus non-nuclear workers at comparable ages.
ences. However, the case numbers are small and only three This phenomenon, which might be termed an “enhanced
dose groups are analyzed which limits any use of trend anal- healthy worker effect”, results from a two step selection pro-
ysis. The small number of deaths can be attributed to the fact cess for nuclear workers. They must be selected to work in
that overhauls began between 1957 and 1967 and recruitment shipyards at time of hire (the usual “healthy worker effect”)
of workers ceased at the end of follow up in 1982 resulting in and then they are selected again to become nuclear workers
only about 13 years as the median period of observation. based on health and work performance while employed.
The population for study represented three groups that Therefore, the analyses of nuclear workers by dose are the
were carefully balanced for several confounding time factors most reliable assessment of the potential risks from low dose
such as age, year of hire and duration of pre-nuclear work radiation in this population. Unfortunately, the lowest dose

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp

 Cancer Mortality in Shipyard Workers 91

of < 5.0 mSv is only analyzed as a category because the and Pierce, D. A. (1995) Lung cancer in radon-exposed min-
sampling strategy was complicated and that group could not ers and estimation of risk from indoor exposure. J. Natl. Can-
be expanded to utilize in dose dependent analysis. cer Inst. 11: 817–827.
The consistency of the finding of an increased risk with 6. BEIR V (1990) National Research Council, U. S. A. Health
Effects of Exposure in Low Levels of Ionizing Radiation,
increasing dose for each cancer in the dose-dependent anal-
National Academy Press, Washington, D. C.
ysis is highly suggestive that radiation is associated with the
7. UNSCEAR (1993) United Nations Scientific Committee.
risk of these cancers. In addition, the LHC cancers, which Sources and Effects of Ionizing Radiation. Report to the General
include leukemias and multiple myelomas as well as non- Assembly with Scientific Annex. United Nations, New York.
Hodgkin lymphomas, are recognized or suspected to be 8. Azzam, E. I., de Toledo, S. M. and Little, J. B. (2001) Direct
associated with an increased risk from radiation exposure. evidence for the participation of gap junction-mediated inter-
These cancers show a significant and very high five to six- cellular communication in the transmission of damage signals
fold excess risk in 50 mSv or more radiation-exposed work- from α-particle irradiated to nonirradiated cells. Proceedings
ers compared to those in the low dose reference group. of National Acad. of Sci. 98: 473–478.
Therefore, these data on increasing risks from radiation 9. Bettega, D., Calzolari, P., Chiorda, G. N. and Tallone-
exposure for LHC support the findings from similar studies Lombardi, L. (1992) Transformation of C3H 10T1/2 cells
with 4.3 MeV α particles at low doses: effects of single and
in the literature. The consistency of the increase in risk with
fractionated doses. Radiat. Res. 131: 66–71.
dose for each cancer and the significantly high risk of LHC
10. Hei, T. K., Wu, L. J., Lui, S. X., Vannais, D., Waldren, C. A.
associated with radiation doses of 10 mSv or more indicate and Randers-Pehrson, G. (1997) Mutagenic effects of a single
that the shipyard population, which was still very early in and exact number of alpha particles in mammalian cells. Proc.
follow up at the time of termination of the study, needs to of the Nat. Acad. Sci. 94: 3765–3770.
have additional follow up to determine the risks of leukemia, 11. Miller, R. C., Marino, S. A., Brenner, D. J., Martin, S. G.,
LHC and solid tumors. These early data suggest the workers Richards, M., Randers-Pehrson, G. and Hall, E. J. (1995) The
may have a risk of some cancers from low dose radiation. A biological effectiveness of radon-progeny alpha particles. II.
precise estimate of risks from nuclear work in this popula- Oncogenic transformation as a function of linear energy trans-
tion could be obtained by examining the mortality for all fer. Radiat. Res. 142: 54–60.
77,000 nuclear workers in the study population plus addi- 12. Miller, R. C., Randers-Pehrson, G., Geard, C. R., Hall, E. J.
and Brenner, D. J. (1999) The oncogenic transforming poten-
tional entrants into the program and extending the follow up
tial of the passage of single α particles through mammalian
to more recent years. The additional data provided could
cell nuclei.” Proc. Natl. Acad. of Sci. USA. 96: 19–22.
identify the boundaries of the estimates of risks from low- 13. Stern, F. B., Waxweiler, R. A., Beaumont, J. J., Lee, S. T.,
level repeated doses of gamma radiation and these results Rinsky, R. A., Zumwalde, R. D., Halperin, W. E., Bierbaum,
could be compared to risks derived from the Japanese Atom- P. J., Landrigan, P. J. and Murray, W. E. Jr. (1986) A case-
ic Bomb survivors who essentially had a single radiation control study of leukemia at a naval nuclear shipyard. Am. J.
dose. Any future analyses also should adjust the radiation of Epidemiol. 123: 980–982.
risks for other job exposures. 14. Kubale, T. L., Daniels, R. D., Yiin, J. H., Couch, J., Schubauer-
Berigan, M. K., Kinnes, G. M., Silver, S. R., Nowlin, S. J. and
REFERENCES Chen, P. H. (2005) A nested case-control study of leukemia
mortality and ionizing radiation at the Portsmouth Naval Ship-
1. Shimizu, Y., Kato, H. and Schull, W. J. (1990) Studies of the yard. Radiat. Res. 164: 810–819.
mortality of A-bomb survivors. Mortality, 1950–1985: Part 2. 15. Matanoski, G. M. (1991) Health Effects of Low-Level radia-
Cancer mortality based on the recently revised doses (DS86) tion in Shipyard Workers, Final Report, DOE/EV/10095-T2,
Radiat. Res. 121: 120–141. National Technical Information Service, Springfield, Virginia,
2. Pierce, D. A., Shimizu, Y., Preston, D. L., Vaeth, M. and U.S.A.
Mabuchi, K. (1996) Studies of the mortality of atomic bomb 16. Steenland, K., Beaumont, J., Spaeth, S., Brown, D., Okun, A.,
survivors. Report 12, Part I. Cancer: 1950–1990. Radiat. Res. Jurcenko, L., Ryan, B., Phillips, S., Roscoe, R. and Stayner,
146: 1–27. L., et al. (1990) New developments in the Life Table Analysis
3. Pierce, D. A. and Preston, D. I. (2000) Radiation-related can- System of the National Institute for Occupational Safety and
cer risks at low doses among atomic bomb survivors. Radiat. Health. J. Occup. Med. 32: 1091–1098.
Res. 154: 178–186. 17. Correa-Villaseñor, A. (1987). A case control study of mesothe-
4. BEIR VI (1988) National Research Council, U. S. A. Health lioma in the shipyard industry. Thesis (Ph. D.)–Johns Hopkins
Risks of Radon and Other Internally Deposited Alpha- University.
Emitters. National Academy Press, Washington, D. C.
5. Lubin, J. H., Boice, J. D., Edling, C.,Hornung, R. W., Howe, Lecture on October 15, 2003
G. R., Kunz, E., Kusiak, R. A., Morrison, H. I., Radford, E. Received on September 15, 2006
P., Samet, J. M., Turmarche, M., Woodward, A., Yao, S. X. Accepted on July 24, 2007
J-STAGE Advance Publication Date: August 10, 2007

J. Radiat. Res., Vol. 49, No. 1 (2008); http://jrr.jstage.jst.go.jp